Re: [femm] Shielding effectiveness

[dcm3c@xxxxxxx]

In a message dated 4/25/01 4:39:26 PM Eastern Daylight Time, 
emiliano.menghi@xxxxxxxxx writes:

> But if i use a second conductor (at distance of 5 meters)
> for the return, the problem is solved?

It all depends on exactly what problem you are trying to solve and what 
results you are trying to obtain. Cases that I'd guess you could be 
interested in are:

1) Currents are conserved in the shield (shield is "floating"), wire carries 
a fixed current. The goal here would probably be to determine the losses in 
the shield. You'd model this setting A=0 at the outer edge of the shield, 
specifying two circuits: one where total current is zero that is applied to 
the shield, and one where the total current is 324 A that is applied to the 
wire. Now, the eddy current distribution in in the shield will be "right", 
so that the losses will come out OK. However, the impedance that the voltage 
reports for the wire won't be that meaningful, because it will be a function 
of where you draw the boundary.

2) Currents are not conserved in the shield, because the shield is 
"grounded". The wire carries a fixed current. The goal here would be to 
find the impedance of the wire as well as the current and losses in the 
shield. And what I mean by "grounded" isn't just grounded at one 
point--this would be the same as "floating" to femm, since femm doesn't do 
capacitive effects. To get the currents not to be conserved within the 
shield, you'd have to have both ends of the shield connected together, so 
that currents can basically make a big "ground loop". In this case, you have 
to model both the the wire and a second conductor for its return. You'd also 
want to model either another shield around the return wire or some other path 
by which currents through the shield complete and electric circuit path. 
There would be two ways that this might be modeled: 
a) if the return wire has the same dimensions as the first wire and has an 
identical shield, you can use an A=0 line of symmetry on line that bisects 
the circuit so that you only have to model one wire. Then, you'd just define 
one circuit property for the model: a 324 A circuit that you'd apply to the 
wire. Because of the symmetry, any current that you induce in the shield 
you'd modeled is returned in the shield that you haven't explicitly modeled, 
so no additional circuit property is necessary. It would be a good idea to 
use some sort of "open" boundary condition on the outer edge here. Multiply 
the impedance that you get from looking at View|Circuit Props to get the 
impedance of the entire wire circuit subject to the shields.
b) if the return for the shield is just something like a wire that you have 
alligator-clipped to each end of the shield, this probably precludes the use 
of symmetry. Then, you'd have to model the wire, its return, the shield, and 
the shield's return, ideally applying some sort of open boundary conditions. 
Then you'd make 3 circuits: one with 324 A and one with -324 A, which are 
applied to the wire and its return respectively, and one with a total of 0A, 
which is applied to both the shield and its return. Sum the impedance 
attributed to the wire and the impedance attributed to its return to get the 
total impedance that you'd have to drive.

3) "floating" shield with the goal of determining the impedance of the wire 
and shield losses. Here, you'd model the shield, the wire, and its return, 
using some sort of open boundary condition for the outer boundary. This is 
the same as 2b), except that the current=0A circuit property is applied to 
just the shield (since there is no shield return wire).

> Hi Dave, i know the problem of non-zero total current and
> for this, i used this configuration:
>
> A wire with R= 7.1 mm
> A dielectric (ext rad = 16.5 mm)
> A tubular shield (ext rad = 19.5 mm)
> A second wire with R= 7.1 mm at 5 m from the first
> A circuit property with I=324 A for wire 1
> A circuit property with I=-324 A for the return (wire 2).
> A=0 at R=25 m
> Wires and shield are in aluminum

> And the question about the length of the wire (and the shield)?

As long as a 2D model is a good assumption, you can just take the 2D 
per-meter results and multiply them by the length of the wires. Where this 
could go wrong is if you are interested in impedances and the distance 
between wires is on the same order as the length of the wires. Then, you'd 
really need a 3D solver to get things straight, but you can probably get the 
results that you are interested in without resorting to this extreme.

Dave Meeker
--
http://members.aol.com/_ht_a/dcm3c